# datalab-to/surya

OCR, layout analysis, reading order, table recognition in 90+ languages

Repository: https://github.com/datalab-to/surya
Canonical: https://ross.abutalabs.com/products/surya
Homepage: https://www.datalab.to
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-08-21T10:00:59+00:00

## Health v2 (maintenance only)
Score: 86/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 98, release rhythm 81, longevity 69
- inputs: {"age_days": 966, "days_push": 12, "days_rel": 44, "gap_med": 3, "n_releases_24m": 62}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 21318, forks 1533 (observed 2026-08-28T04:11:31.583882+00:00)

## What it is
Surya is a 650M parameter OCR toolkit from Datalab providing state-of-the-art text recognition, layout analysis, reading order detection, and table recognition in 90+ languages. It ships as a Python library with GPU acceleration and is part of Datalab's document intelligence model suite.

## Use cases
- extract text from scanned pdfs
- ocr documents in multiple languages
- detect layout regions and reading order in documents
- recognize tables from images
- convert scanned documents to markdown
- run ocr on gpu at high throughput

## When to choose
- you need multilingual ocr across 90+ languages
- you need layout analysis and table recognition alongside text extraction
- you want a self-hosted, Apache-2.0 licensed ocr pipeline with gpu speed
- you are building document processing or rag ingestion pipelines

## When to avoid
- you only need simple English-only ocr with a lightweight dependency
- you have no gpu and need very fast batch processing
- you need a fully managed api without running models yourself
- your models must be under a permissive license without usage restrictions (weights use OpenRAIL-M)

## Facets
- artifact type: library
- maturity: active
- function: ocr, image-processing, machine-learning, deep-learning, pdf
- domain: computer-vision, pdf, machine-learning, files
- platform: python, cross-platform
- tags: document-intelligence, layout-analysis, table-recognition, reading-order, multilingual, text-detection, natural-language-processing, gpu

## Member repositories
- datalab-to/surya (main) score 86

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:31.583882+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T16:58:24.914009+00:00, confidence not recorded.
  - readme: https://github.com/datalab-to/surya (fetched 2026-08-28T04:11:31.583882+00:00, sha 69b053d2917a)
  - homepage: https://www.datalab.to (fetched 2026-08-29T07:56:49.830598+00:00, sha 2bbd3c9c83e2)
- Data as of 2026-08-30T08:39:29.467469+00:00.